About Me

My name is Fangkai Jiao (焦方锴). I am a third-year PhD candidate at Nanyang Technological University and Institute of Infocomm Research (I2R), A*STAR, with A*STAR SINGA Scholarship. I am fortunate to be supervised by Prof. Shafiq Joty, Prof. Nancy F. Chen and Prof. Aixin Sun. Before that, I obtained both my M.Eng. and B.Eng. degree from Shandong University in 2022 and 2019, respecitively, under the supervision of Prof. Liqiang Nie. I also work closely with Dr. Yangyang Guo, Dr. Zhiyang Teng and Zhengyuan Liu. I have interned at CoAI Lab, Tsinghua University, mentored by Prof. Minlie Huang, DAMO Academy, Alibaba-inc, mentored by Dr. Feng Ji and Dr. Feng-Lin Li, and Langboat Technology.

My primary research interest includes self/weak-supervised training and machine reasoning with large language models. I’m also one of the main contributors to PandaLLM project.

News

11/14/2024: Our paper Learning Planning-based Reasoning by Trajectories Collection and Process Reward Synthesizing has received the Outstanding Paper Award at EMNLP 2024.

09/20/2024: Learning Planning-based Reasoning (Process-DPO) is accepted by EMNLP 2024.

07/26/2024: UNK-VQA is accepted by TPAMI.

05/15/2024: Joined MSRA as an intern.

04/01/2024: Our survey and the integrated LLM contamination detection tool is released. Check the paper and Github repository.

03/14/2024: Two papers are accepted by NAACL 2024.

10/08/2023: Our survey is accepted by Findings of EMNLP 2023.

08/10/2023: Weights of Panda-Llama2-13B & Panda-Llama2-13B-Chat are released. And PandaLLM project is also upgraded to PandaLLMOps. Check our tutorial about large scale distributed training here.

06/24/2023: PandaLLM-13B-Chat weights are released. Check it here.

06/13/2023: PandaLLM-13B-Chat on telegram bot is released. Check it at our repo.

05/14/2023: PandaLLM-13B is released.

05/02/2023: Our Chinese LLM - Panda LLM is released with 7B weights. Better reasoning capability. Check it here!

03/15/2023: Our survey about mutli-modal retrieval augmented generation is uploaded to arxiv.

10/05/2022: One paper is accepted by IEEE/ACM Transactions of Audio, Speech and Language Processing (TASL).

03/31/2022: One paper is accepted by SIGIR 2022.

02/24/2022: One paper is accepted by Findings of ACL 2022.

02/09/2022: Our system, MERIt (Deberta-v2-xxlarge) achieves a new SOTA on Reclor leaderboard.

11/05/2021: Our system for logical reasoning achieves the Top-1 results among single models on Reclor leaderboard.

Academic Services

Reviewer

Confernece: ACL Rolling Review, EMNLP 2023, ACM MM 2022 (workshop), ICLR 2024

Journal: IEEE/ACM TASL, IEEE TKDE, IEEE TMM